PulPy: Pulses in Python

Source Code | Documentation | Tutorials

Description

Welcome to PulPy! PulPy is a Python package for RF pulse and gradient pulse design for MRI. PulPy focuses on the design of individual pulses rather than full pulse sequences (for this, see other packages, e.g. PyPulSeq).

PulPy is the successor package to SigPy.RF, a sub package for RF pulse design nested inside of the SigPy package for signal processing and image reconstruction. PulPy is designed to replicate functionality of SigPy.RF, but is able to stand alone as an independent Python package. It also places greater emphasis on MRI gradient waveform design, which is necessary for many classes of RF pulses but also has broader applicability. PulPy current relies on SigPy as a dependency and takes advantage of its' abstraction classes. This dependency will be reduced in future releases.

Installation

PulPy requires Python version >= 3.8. The core module depends on numba, numpy, sigpy, PyWavelets, scipy, and tqdm.

Via pip

PulPy can be installed through pip:

pip install pulpy

Developer Installation

If you want to contribute to the PulPy source code, we recommend you install it with pip in editable mode:

cd /path/to/pulpy
pip install -e .

To run tests and contribute, we recommend installing the following packages:

pip install coverage ruff sphinx sphinx_rtd_theme isort

and run the script run_tests.sh.

If you make modifications and would like to update the version number across the release, run bumpversion (where argument size is one of 'major', 'minor', 'patch'):

bumpversion --current-version 1.8.2 size

Getting Started

To begin using pulpy, import the package in your Python script. For demo purposes, we'll also import the SigPy package's plotting functions:

import pulpy as pp       # import full package
import sigpy.plot as pl      # import a plotting function

1) RF Pulse Design and Simulation

pulpy allows you to easily design and simulate RF pulses. Here's an example of an SLR pulse design with multibanding:

1a) define RF pulse parameters

tb = 8      # RF pulse time-bandwidth product
N = 512     # number of timepoints to design
d1 = 0.01   # magnetization passband ripple level
d2 = 0.01   # magnetization stopband ripple level
p_type = 'ex'   # RF pulse type - a 90 degree excitation pulse
f_type = 'ls'   # filter type for SLR design - using a least squares filter

1b) design and plot the RF pulse

pulse = pp.rf.slr.dzrf(N, tb, p_type, f_type, d1, d2, True)
pl.LinePlot(pulse,mode='r')     # plot the real component of the RF pulse

1c) multiband the single-band RF pulse to excite multiple slices simultaneously

n_bands = 3              # design to excite 3 bands of magnetizaztion
phs_type = 'phs_mod'     # 'phsMod', 'ampMod', or 'quadMod' - the method of designing the pulse phases
band_sep = 5*tb          # separate by 5 slice widths
mb_pulse = pp.rf.multiband.mb_rf(pulse, n_bands, band_sep, phs_type)
pl.LinePlot(mb_pulse)

1d) simulate the transverse magnetization profile of both pulses. We do this by first calculating the Cayley-Klein parameters representing the rotation of the magnetization vector produced by the RF pulse (variables 'a' and 'b'). We then use the relationships in Pauly et. al. to convert this to the resulting excitation magnetization.

[a, b] = pp.sim.abrm(pulse, np.arange(-20*tb, 20*tb, 40*tb/2000), True)
Mxy_single_band = 2*np.multiply(np.conj(a), b)  # from Pauly et. al. IEEE TMI (1991). 
[a, b] = pp.sim.abrm(mb_pulse, np.arange(-20*tb, 20*tb, 40*tb/2000), True)
Mxy_multi_band = 2*np.multiply(np.conj(a), b)  # from Pauly et. al. IEEE TMI (1991). 
pl.LinePlot(Mxy_single_band, title='single band excitation')
pl.LinePlot(Mxy_multi_band, title='multi-band excitation')

1e) Export the RF pulse to GE format for use in a scanner. We will compute the important parameters then write to .i file:

pp.ge_rf_params(pulse, dt=4e-6)   # prints out the most important GE parameters
pp.signa(pulse, 'slr_ex')         # writes to .i file

2) Gradient Waveform Design and Optimization

pulpy also has a variety of tools for designing gradient pulses. This ranges from simple trapezoids, the building block of many pulse sequences:

dt = 4e-6  # s
area = 200 * dt
dgdt = 18000  # g/cm/s
gmax = 2  # g/cm

trap, _ = pp.grad.waveform.trap_grad(area, gmax, dgdt, dt)

pl.LinePlot(trap, title='trapezoidal gradient')

to more complex time-varying waveforms (e.g. spiral gradient waveform):

fov = 0.55    # imaging field of view [m]
gts = 6.4e-6  # hardware dwell time [s]
gslew = 190   # max. slew rate [mT/m/ms]
gamp = 40     # max. amplitude [mT/m]
R = 1         # degree of undersampling
dx = 0.025    # resolution

# construct a trajectory
g, k, t, s = pp.grad.waveform.spiral_arch(fov / R, dx, gts, gslew, gamp)

pl.LinePlot(np.transpose(g),mode='r', title='spiral gradient (1 axis plotted)')

to a few tools for more advanced design (e.g. min-time-gradient designers, which modifies an existing trajectory to be time-efficient):

import math        

t = np.linspace(0, 1, 1000)
kx = np.sin(2.0 * math.pi * t)
ky = np.cos(2.0 * math.pi * t)
kz = t
k = np.stack((kx, ky, kz), axis=-1)

(g, k, s, t) = pp.grad.optim.min_time_gradient(
    k, 0.0, 0.0, gmax=4, smax=15, dt=4e-3, gamma=4.257
)

Documentation

Documentation for PulPy is available at ReadTheDocs.

A series of Jupyter notebooks have been developed that provide tutorials of several classes of pulse design at the demo code repository. Simply clone this repository, install Pulpy (and Jupyter notebook), and get started designing pulses!

Contact and Contribution

We welcome feedback on this project! It is a work in project, so please report bugs and issues on GitHub. We also encourage you to contribute additional pulse design tools. Point of contact: jonathan.bach.martin@vumc.org.